I have configured the assistant to use OpenAI Real-time API. Worked perfectly yesterday but this morning, when I had a customer demo, the assistant was not picking up information from the system prompt at all. Had to switch to GPT4o cluster. Any ideas why that is happening?

@NiXLiM I'm sorry to hear that you encountered issues during your customer demo. There could be several reasons why the assistant was not picking up information from the system prompt, ranging from temporary service disruptions to configuration changes. Here are a few steps you can take to troubleshoot the issue: 1. **Check Service Status**: Verify if there were any known outages or service disruptions with the OpenAI Real-time API at the time of your demo. 2. **Review Configuration**: Ensure that the assistant's configuration has not been altered since the last successful operation. This includes checking API keys, endpoint URLs, and any other relevant settings. 3. **Test Connectivity**: Try a simple test to see if the assistant can currently connect to the OpenAI Real-time API and receive responses. 4. **Review Logs**: If possible, check the logs for any error messages or warnings that occurred during the time of the demo. 5. **Contact Support**: If the issue persists and you cannot find a resolution, reaching out to OpenAI support may provide more insight. If you'd like, I can assist you in closing this support ticket if the issue has been resolved, or if you need to perform any of the above actions, please let me know how I can assist you further.

Hey @NiXLiM can you share the call id?

Hi, are you still having problems with the tool triggering? I'm experiencing the same issue—it sometimes works, and other times it doesn't... I'm using gpt4 cluster

Yep, having the issue still @cs

Hey @NiXLiM it's using your system prompt which you can check through prompt the issue here is voice not voicing out the words correctly and which is resulting into bad description. As it's voice to voice model so you have to bear unknowns intially.

transcript

AI: Hi. This is Muscle Bassin Jim Alloy speaking. How can I help you?\nUser: Uh, what's the address of the gym, please?\nAI: Sure. Could you let me know which gym you're looking for the address of? There are many gyms out. Sure. Could you let me know\nUser: This 1 is\

@Shubham Bajaj Hey! Thank you for getting back to me. This does not make sense though - I did not change the prompt from the day it was working flawlessly to the day it stopped. I had to use Retell for the demo yesterday, which is disappointing for me personally cuz I am not a fan of Retell, but it was working. What do I need to do for the VAPI integration to work?

Hey @NiXLiM:- 1. I never mentioned you changed your prompt 2. Voice to Voice Model is not working as expected 3. Voice coming out of Realtime model is not accurate because of this your observing transcription issue 4. It will work with Gpt-4o well 5. I am taking your feedback and sharing with team, regarding Realtime Api Voice. 6. Lastly, voice to voice model is still new so sometimes these issues are expected to exist.